Skip to content

How brand equity is measured, and why the method matters.

Any language model will score a brand if you ask. Run it twice and the number changes. Show it a famous brand and the number rises — because the model knows the brand, not because the brand is better. Smart Branding was built so that does not happen: the score comes from evidence, not opinion.

Why a brand score needs a method.

A brand score is only worth something if it is three things at once, and most machine-generated scores are none of them.

  • Defensible

    There is a theory of brand behind it, not a statistical guess.

  • Reproducible

    The same brand, with the same evidence, gets the same score today and three months from now.

  • Actionable

    It breaks down into criteria that tell you what to do to raise it.

The core of the method is an inversion: the intelligence does not judge, it collects — the rubric judges. The model extracts observable evidence about the brand’s presence; a deterministic formula turns that evidence into a score.

Four traditions in brand equity, and what we take from each.

Smart Branding does not start from zero. It declares what it inherits and what it leaves behind from four bodies of literature.

From Aaker
The founding premise: a brand is an economic asset with measurable dimensions. What we leave behind is the dependence on large-scale declared research — our unit of analysis is the brand’s public evidence, not the survey.
From Keller and the CBBE model
The layered architecture: a brand is built from the ground up, and its value lives in the mind of the public.
From Sharp and Romaniuk
The most important correction of the last two decades: being recognisable matters as much as being different. Classic literature measured only the second.
From Kapferer
The principle that identity precedes image: the brand declares who it is, and the market returns a reflection. The distance between the two is what we call consistency.

What the four have in common is that none of them needs a machine’s opinion. All define a strong brand by observable behaviour, and observable behaviour is what a rubric can verify.

The three axes: singularity, consistency and positioning.

Every brand diagnosis answers three questions, and each one is observable in public evidence without primary research.

  • Singularity

    Cover the logo — is the brand still recognisable? Put it beside its five closest competitors — is it replaceable? Singularity is the synthesis of two ideas the literature usually keeps apart: distinctiveness, being recognisable, and differentiation, being different. A singular brand is both.

  • Consistency

    Is the brand that declares itself the same one that shows up, everywhere it shows up? This is the most auditable of the three, and it is a condition for the other two: singularity without consistency does not build memory, and positioning without consistency does not build meaning.

  • Positioning

    In one sentence: what place does this brand claim, for whom, against which alternative? And does that sentence hold up in what the brand actually shows?

To be verified

Final wording of the three axes depends on the construct definition under way.

Not for launch — paginas-internas.md

How a brand score is built, step by step.

The score travels through four stages, and the intelligence only takes part in the first.

  1. 01

    Collection

    The model scans the sources in scope and answers closed factual questions. Never “is this brand consistent?” — always “do the Instagram and the site use the same register? cite three examples of each.” Every piece of evidence carries its source, and evidence without a source is discarded.

  2. 02

    Verification

    Each axis has five fixed criteria, checked against behavioural anchors at three levels: absent, partial, full. The anchors describe observable facts, not qualities.

  3. 03

    Calculation

    The axis score is the sum of its five criteria, from 0 to 10. No weighting, no discretionary adjustment. The same evidence always produces the same score, by construction.

  4. 04

    Reading

    The score falls into a band with a fixed meaning, and every criterion below full automatically becomes a recommendation — because the anchor for full describes exactly what is missing.

The bands

0–3
Fragile
4–6
Developing
7–8
Solid
9–10
Reference

To be verified

Weights, minimum coverage and stabilisation across passes depend on the definition under way.

Not for launch — paginas-internas.md

Bias in AI brand evaluation is a finite list.

Every known bias in machine evaluation has a structural countermeasure in the method. Bias is not a mystery to be apologised for — it is a list of engineering problems.

  • Familiarity

    A famous brand scores higher because the model knows it.

    Countermeasure

    Only evidence collected in scope, with a citation, can reach the score. What the model already knows is inadmissible.

  • Size

    High output looks like a strong brand.

    Countermeasure

    Every criterion is written as a quality of presence, never a quantity. Followers, posts and media volume appear in no anchor.

  • Variance

    The same brand, different scores.

    Countermeasure

    The score is calculated by formula over discrete criteria, never produced as a number by the model.

  • Compliance

    The model’s tendency to please.

    Countermeasure

    The model never sees the question “what score does this deserve?”. Where no judgement is delegated, there is nothing to please.

  • Position and verbosity

    Order and volume of evidence sway the result.

    Countermeasure

    Criteria are evaluated one at a time, in isolation.

  • Market

    Global yardsticks applied to local context.

    Countermeasure

    Reference competitors are mapped inside the brand’s real market.

  • Halo

    One strong axis contaminates the others.

    Countermeasure

    The three axes are extracted and calculated independently.

The rubric is versioned like software.

Method governance is what keeps a score comparable over time. Every score records the rubric version that produced it, and anchors change only in a new version, never retroactively.

Before a version ships, it runs against a fixed set of calibration brands with an established human consensus. If the machine diverges beyond tolerance, the version does not ship.

The same principle that governs the product governs the method: evolution is measured, not promised.

See the method run on your brand.

The method runs inside the brand intelligence, and what it finds in each sector is published as sector research.

Unlock my BR4NDCODE